Euclidean geometry defines objects that can be realized in space, and may therefore be founded in spatial perception. We investigated whether the perception of small, 2-dimen- sional visual forms could provide cognitive foundations for Euclidean knowledge, by asking two questions. First, are humans sensitive to form variations that are relevant to Eu- clidean geometry (e.g. changes in angle)? Second, can ob- servers easily disregard variations that are irrelevant to Eu- clidean geometry (e.g., changes in scale)? Participants from the U.S. (age 3-34 years) and from the Amazon (age 5-67) were asked to locate deviants in panels of 6 forms of vari- able orientation. Results indicate that perception of forms aligns with a restricted version of Euclidean geometry, where forms are defined in terms of metric proportions and global size, but mirror images are assimilated. Moreo- ver, children below 6 did not clearly analyze forms in terms of the shape property of angle.